Scanner Pro does the OCR - and it was implied (due to the need for the correct name) that the categorisation is a mapping from the name used for the scanned OCRed file.

That is my take from it. I could be wrong.